AI therapy is mental health support given by software rather than a person. That covers specialised mental health apps, AI chatbots and the general-purpose chatbots many people now turn to when they want to talk something through. These tools may offer emotional support, mood tracking, journalling prompts, plain-English explanations of how the mind works, and exercises based on approaches such as cognitive behavioural therapy (CBT).
In our view, an AI tool is not a therapist. It can help you reflect or get through a hard evening, but it is not qualified, it cannot give dependable crisis care and it should not replace professional support when you need it. You may well find it convenient and useful. We would just like you to understand its limits before you share anything sensitive or rely on its advice.
What AI therapy means, and why it is not quite therapy
“AI therapy” is a loose umbrella term. It is not a single clinical service with agreed standards, and “AI-driven mental health support” or “AI coaching” describes it more accurately. Counselling and psychotherapy are delivered by a trained person who is usually accountable to a professional body, works within a code of ethics and has regular clinical supervision. An app can borrow the language of therapy without any of that structure behind it.
When people talk about AI therapy, they usually mean one of these:
- dedicated mental health chatbots built around a structured approach
- mood and symptom trackers that look for patterns over time
- guided journalling apps that respond to what you write
- AI companions designed for ongoing conversation
- self-help apps that tailor exercises to you
- general chatbots, the kind many of us already use for work, asked for emotional support
People use them to put feelings into words, look back on a difficult conversation with a partner or manager, notice a thought pattern that keeps coming back, learn a grounding exercise, track their mood, or prepare questions for a therapist or GP.
You may also hear about AI being used behind the scenes in health services for notes, scheduling or planning. That is a separate thing from a tool you talk to directly. Among the apps you talk to, the label matters less than the design. An app that calls itself a wellness tool may offer something very different from one built around a recognised therapeutic framework, so we would not treat any two as interchangeable.
How AI mental health chatbots work
Using an AI chatbot usually feels like sending messages. You type or speak about what is happening, and the system analyses your words and replies with a response, a question, a prompt to reflect or a suggested exercise. If memory is switched on, it may refer back to earlier conversations.
Most newer tools run on large language models, a type of artificial intelligence trained on huge amounts of text. Using natural language processing, they read your message and predict a reply that sounds helpful, based on patterns in that text. They do not understand you, care about you or form a therapeutic relationship in the human sense, even when a reply feels personal.
Some tools add machine learning and sentiment analysis. In practice, that means they look for words linked to distress, or they adjust their suggestions based on what you have shared before. These signals can help, but they can also be incomplete or wrong. A tool may miss distress written in calm language, or treat an ordinary bad day as an emergency.
Specialised mental health tools and general-purpose chatbots are also very different. A dedicated tool may follow structured, CBT-style prompts and built-in safety rules. A general chatbot was not necessarily designed or tested for mental health support at all.
Warm, confident and empathetic language can be very convincing. Keep in mind that the tone of a reply is no evidence that what it says is accurate.
What AI support may help with
The appeal is easy to see. An AI tool is there at 2am when you are lying awake with your heart racing. The people around you do not see it, it is often free or cheap to start, and it lets you write your thoughts down when saying them aloud to someone feels like too much.
The lower-risk uses are the practical ones: guided breathing, grounding, journalling, mood tracking, habit reminders, basic information about anxiety or sleep, and practising how to raise a concern with a partner, your GP or a therapist.
For everyday stress, mild anxiety or low mood, some of the research is encouraging. A 2023 meta-analysis in npj Digital Medicine pooled 15 randomised trials and found that AI conversational agents reduced symptoms of depression and distress, although the effect on overall wellbeing was not significant. In an early randomised trial of a generative AI therapy chatbot, researchers at Dartmouth reported falls in depression and anxiety symptoms over eight weeks. They also said that this kind of tool still needs clinician oversight. Evidence for one carefully built and supervised programme does not carry over to every chatbot in the app store. In the UK, the NICE evidence standards framework for digital health technologies sets out the evidence these tools should be able to show, with more demanding standards for tools that do more. That is a sensible test to apply to anything you use.
We think AI fits best as one part of a wider plan: between counselling sessions, while you wait for an appointment, or as a prompt to reflect, alongside sleep, relationships, community and professional care. Its biggest benefit is that it is always available, and that can make a small first step easier. But being there at any hour does not make a chatbot clinically responsible for your care.
The risks of using AI for mental health support
AI can get things wrong, and it rarely tells you when it has. It can misunderstand context, repeat biases from the text it learned from, give advice that does not suit your situation, or present a made-up answer with complete confidence. The World Health Organisation has warned that large language models can give health answers that sound authoritative but contain serious errors, and it has called for transparency, rigorous evaluation and human oversight. For that reason, a chatbot should never diagnose a mental health condition, assess your medication or tell you to change your treatment.
A chatbot also has no human judgement. In a session, a therapist notices the pause before you answer, how you seem different from one week to the next, what is going on in your family, a safeguarding concern, or the thing you keep circling but cannot quite say. A chatbot has only your words, and it cannot reliably read any of that.
False reassurance is a quieter risk. Many chatbots are built to be agreeable, so they may keep agreeing with you when a gentle challenge would help more. If you turn to a chatbot every time things feel hard, it can delay you speaking to a GP, a therapist or someone you trust, especially if your symptoms are getting worse.
AI is not a safe crisis service. If you are having suicidal thoughts, thinking about self-harm, experiencing psychosis, severe depression or mania, living with abuse, struggling with an eating disorder or in immediate danger, you need a person. Pattern detection and automated crisis prompts can fail, and a missed warning sign matters most at exactly these moments.
Some people find comfort in an AI companion, and that is understandable. Something that always replies, never tires and never judges can feel like a relief. The concern is when it starts to feel like the only place you are understood, or when time with it replaces time with people. If you notice that happening, it deserves your attention, and there is no shame in it. We have written more about this pull in wired for connection, drowning in notifications.
Replies may also not take account of your age, disability, neurodivergence, culture, sexuality or religion, or of what life in Ireland is like. Advice based on another country’s health system, or on assumptions about family and work that do not match yours, can leave you feeling more alone rather than less.
Privacy and regulation of AI chatbots in Ireland
Typing something into a chatbot is not the same as speaking in confidence with a therapist. Depending on the product and its terms, your conversations may be stored, read by staff, shared with service providers, used to improve the system or transferred outside Ireland and the EU.
Before you share anything personal, we would suggest you:
- read the privacy notice, especially any section on health information
- check whether your conversations are used to train the model, and opt out if you can
- look for settings that control how long data is kept and how to delete it
- switch memory off if you do not want earlier conversations kept
- use a strong password that you do not use anywhere else
- leave out names, addresses, workplaces, medical record details and anything that identifies another person
The General Data Protection Regulation (GDPR) gives you rights when an organisation processes your personal data. The Data Protection Commission sets these out, including the right to see your data, to have it corrected and, in some circumstances, to have it erased. Those rights matter, but they do not mean a chatbot is clinically confidential, secure or suitable for your most sensitive disclosures.
No single approval system covers everything marketed as AI therapy. The Health Products Regulatory Authority notes that software can meet the definition of a medical device, but that a medical device must have a medical purpose. A general wellness chatbot may fall outside those rules entirely.
The standards that apply to qualified counsellors, psychotherapists and psychologists, such as training, ethics, supervision and complaints procedures, do not automatically apply to a chatbot. If you are looking for a human practitioner, you can check whether they are accredited by or affiliated with a professional body. These include the Irish Association for Counselling and Psychotherapy (IACP), the Irish Association of Humanistic and Integrative Psychotherapy (IAHIP), the Irish Council for Psychotherapy (ICP) and the Psychological Society of Ireland (PSI).
Using AI more safely alongside real support
If you do use an AI tool, a few habits help keep it in its place:
- use it for reflection, not diagnosis
- check any health claim against a reliable source such as the HSE
- never change your medication or treatment because a chatbot suggested it
- keep sensitive personal details to a minimum
- notice whether you come away calmer, more distressed or more isolated
If a reply feels unsafe, shaming, extreme or simply wrong, stop following it. Save the text if that helps, report it through the platform, and check the advice with your GP, a qualified mental health professional or someone you trust.
If AI is shaping your decisions or how you feel day to day, tell your therapist. We would be glad to talk about it in a session with us. Together, we can look at what you get from it and where firmer boundaries or different support might serve you better.
You do not need to wait for a crisis, or prove that your worries are serious enough, before you talk to a person. Human support matters most when distress lingers, daily life is getting harder, relationships are suffering or symptoms are building. If you are unsure, our guide to knowing when you might need therapy may help.
If you may harm yourself or someone else, call 112 or 999 or go to your nearest emergency department. You can also call Samaritans free on 116 123. The HSE urgent and emergency care page explains where to get immediate help.
FAQs
Is AI therapy suitable for everyone?
No. It depends on the person, the tool and how much support you need. Children and teenagers, anyone in crisis and anyone whose symptoms are severe or getting worse quickly need a person involved, not a chatbot on its own. For young people, our counselling and psychotherapy for teenagers is one place to start.
Can AI therapy replace a qualified human therapist?
No. It may add to your care, but it cannot offer the judgement, accountability, relationship or safeguarding of a qualified professional. We compare the two side by side in our article on AI therapy versus human therapy.
How much does AI therapy cost, and are free tools safe?
Tools range from free versions to monthly subscriptions. The price tells you nothing about safety or quality, and free services may come with limits, advertising or different ways of handling your data. Before you use any tool, look at its evidence, its privacy terms and what it does if you tell it you are in danger.
Talking to a person whenever you are ready
You do not need to know exactly what kind of support you need before you get in touch. Our team offers counselling and psychotherapy in Dublin and Galway, and online sessions anywhere in Ireland. Our current fees are on our fees page, and we keep a low-cost counselling option for people who would otherwise go without. Whenever you feel ready, you can book online or contact us with any questions.